Total Agents
21
A1 – A21
MVP Complete
11
A1 – A11
Scaffold (Phase 3)
10
A12 – A21
Total Runs
—
Awaiting API
API
Offline
Run start_api.sh
Architecture
v3
3 layers · 21 agents
Agent Status by Layer
Start the API to see live status.
Recent Runs
No runs yet.
Quick Start
① Start the Agent API
cd agents/
bash start_api.sh
# or manually:
uvicorn api.main:app --reload
② Connect Dashboard
Default API URL: http://localhost:8000
The green dot confirms connection.
Dashboard polls every 30s automatically.
The green dot confirms connection.
Dashboard polls every 30s automatically.
③ Test Agents
Agents tab → Run Test on any card.
Test Console → Custom JSON payloads.
Run History → Full audit log.
Test Console → Custom JSON payloads.
Run History → Full audit log.
Building agent registry…
System Architecture — net könyvelő v3
From ARCHITECTURE_v3.md
Layer Explanations
Data Flow — Key Facts
BaseAgent — all 21 agents inherit shared/base_agent.py. Every run writes to
agent_run and audit_event tables automatically.Storage (MVP) — per-agent SQLite files under
A{N}/data/a{n}.sqlite. Cross-agent reads use path references. Migration path → shared PostgreSQL.Test isolation — each agent has its own Streamlit app (port 850N) for unit testing. This dashboard calls the API to run all agents from one place.
Human escalation — any agent calls
post_review_task() → inserts into A5's workflow_task table. A5 is the single review queue for the whole system.Compliance chain — A8 monitors legislation changes → feeds A9 (country rules) → A9 is consumed by A2, A4, A14, A18. All hard-coded country values must route through A9.
Cross-Agent Contracts (v3)
These are the stable integration points. Treat as APIs — don't break them across agent versions.
Run History — All Agents
No runs yet — start the API and run a test.
Test Console — Send Custom Payloads to Any Agent
Pre-filled with working test data per agent.
API must be running locally to execute tests.
API must be running locally to execute tests.
OUTPUT
// Select an agent above, then click ▶ Run Test